In Part I it is pointed out that only a few of the existing 50 radio-telephone links are at present working at anything like full-load capacity. For this reason the chief problem for the engineers on most of the links is one of reducing operating costs. On a few heavily-loaded links, however, the expense of some improvement in circuit performance would be justified. Various means are discussed in outline for dealing with these two problems.In Part II further details of the improvements proposed in Part I are dealt with, and in particular the probable value of the single side-band method in relation to other possible solutions is discussed.(1)Operating costs.—It is found that on an existing typical link, with either light or heavy traffic, the single side-band suppressed-carrier system would be expected to save about 87 per cent of the valve replacement costs, and 90 per cent of the cost of the power.(2)Improved performance.—In the cost estimates of (1) it is shown that changing to single side-band working should be equivalent in. signal/noise ratio to increasing the peak power of the transmitter about 16 times. If this factor is allowed for, by greatly reducing the transmitter size, it is shown that single side-band working should still give circuit improvements in relation to fading, quality, and privacy. The single side-band system eliminates the chief cause of distortion at present, i.e. that due to selective fading-out of the carrier. It should also make workable several well-known privacy schemes at present useless during selective fading, and give rise to a promising new method.A possibly new means of reducing distortion due to intermodulation in Class 3 single side-band amplifiers is referred to.In Part III an attempt is made to answer the following three questions: 6pIs the problem technically possible at all under present commercial conditions?If it seems possible, what is likely to be the best general method—independent stable oscillators at each end, or the use of a pilot signal?Having decided on the most promising general method, what problems will be met in designing practical equipment, and how can we solve these problems?The answer to the first is in the affirmative; single sideband working should be possible in practice, either by separate stable oscillators or by a pilot signal.6pA discussion of the second question results in favour of the pilot method, particularly as it is shown that some sort of pilot is almost indispensable in any case, for controlling the medium-period fading.6pAnswering the third, it is suggested that a single radio frequency corresponding to an audio frequency of about 3.4 kilocycles per sec. would be most suitable, and that this pilot should control the frequency of a local oscillator. It is pointed out that exact synchronizing in the correct phase would in any case be useless at present; while selective fading is still not overcome, the exact signal wave-form can never be reproduced at the receiver, even with perfect carrier synchronism.6pExperiments were carried out by the author on three shortwave links using equipment on the lines suggested in Part III, and the results are given in Part IV. The final tests between Madrid and Paris showed that the expected improvements-over double side-band working, discussed in Part II, were fully obtained. A maximum synchronizing error of 6 cycles per sec. over periods of several hours was found.6pIn Part V it is suggested that—6p(1) A simplified form of the receiver described in Part IV could be used commercially, and would meet requirements for at least the next few years. In this connection a somewhat new form of filter, on the “balanced reaction” principle, is described.6p(2) For the immediate future the principle used in the transmitter should be that of the “side-band balance”, i.e. a method of selecting the side-band in one stage directly at the final high frequency, without relying on filters. This circuit is discussed.6pAssuming selective fading to be absent, it is shown in Part VI that synchronizing at the exact frequency, within predetermined phase limits, is fairly simple. One form of circuit to do this is described.6pA high-quality circuit is explained, giving exact synchronism in phase except during periods of deep fading, when the frequency difference may amount to a few cycles per second.